>  Why does the data for a CDS need to be on a doubleword anyway?

"It just does"
Or if you want the more gory explanation... it has to do with storage operand 
consistency rules. CDS is built on top of the hardware memory architecture and 
so it has to play within the same rules as other storage access operations. The 
hardware guarantees consistent (i.e. the same to all observers) block 
concurrent access to the contents of a doubleword but if you cross a doubleword 
boundary you are by definitiion accessing two different double words and you 
lose the consistency guarantee. Read about it in Principles of Operation.

CC                                        

Reply via email to